Actions
Forum Thread Details
  • Replies 5 replies
  • Subscribers 58 subscribers
  • Views 774 views
  • Users 0 members are here
Related

Changes on BBB switching to AM3358BZCZ100

morgaine

CircuitCo's answer in the BeagleBone_Black_FAQ states that when future BBBs switch their SoC to the AM3358BZCZ100 variant, the ability to use EtherCAT (an extremely low latency fieldbus that can also carry normal Ethernet and IP frames transparently) will be lost.

 

Although this isn't likely to affect many people, it's always a pity when some hardware capability is lost, and so I'd like to understand more precisely how the AM3358BZCZ100 differs to make EtherCAT no longer possible on future BBBs.  I'd be grateful for any detailed hardware explanation about this.

 

Morgaine.

  • From my limited understanding of EtherCAT, there doesn't seem to be any point to it unless you have two Ethernet ports so that your device can act as a low-latency pass-through Ethernet switch.  Do you know if you can use the second AM3358/9 Ethernet port on BBone?

  • in reply to johnbeetem

    You can construct EtherCAT systems using single-NIC hosts and clients by implementing the routing fabric in switches set to replicate on all ports.  All the usual physical topologies are possible, line, tree, star, etc.  Not sure about ring, but the common 2-NIC multidrop style is just one implementation of EtherCAT routing, and it's somewhat orthogonal to what is essentially point-to-point messaging.

     

    Addendum. Although my question was really just to understand properly why EtherCAT capability is lost, it's also worth saying that EtherCAT metrics are pretty awesome and really use this comms medium to its full potential. The SoC change is probably intended to save just a few cents per board, which can be important to CircuitCo when margins are so tiny in the "Pi era", but I hope they will continue to provide a few EtherCAT-capable boards at slightly higher prices as well.  It's just too nice to lose entirely.

  • in reply to morgaine

    Morgaine Dinova wrote:

     

    Addendum. Although my question was really just to understand properly why EtherCAT capability is lost, it's also worth saying that EtherCAT metrics are pretty awesome and really use this comms medium to its full potential. The SoC change is probably intended to save just a few cents per board, which can be important to CircuitCo when margins are so tiny in the "Pi era", but I hope they will continue to provide a few EtherCAT-capable boards at slightly higher prices as well.  It's just too nice to lose entirely.

    Well, at ti.com the quantity 1K budgetary pricing for otherwise identical AM3358/9 parts differ by US$2.25 to US$2.34, roughly 10% of quantity 1K pricing.  Here's my wild guess: the chips both have EtherCAT, but it's disabled on AM3358.  Why?  Maybe because TI has to pay fairly expensive patent royalties on chips that have it enabled.

     

    Just a guess -- IANAL.  OTOH, I have observed that whenever good tech ideas are locked away from end users, tassel loafers are not far away image

  • in reply to johnbeetem

    John Beetem wrote:

     

    Just a guess -- IANAL

     

    At the totally acceptable risk of annoying the parasites of a technological civilization, the right acronym is IANAFL.

  • in reply to morgaine

    Morgaine Dinova wrote:

     

    John Beetem wrote:

     

    Just a guess -- IANAL

     

    At the totally acceptable risk of annoying the parasites of a technological civilization, the right acronym is IANAFL.

    Aw c'mon, they're not all bad.  It's just the 90% who make the other 10% look bad [from the Old Jokes Home image ]